A Novel Spatial-Temporal Regularized Tensor Completion Algorithm for Traffic Data Imputation

Abstract

In the Intelligent Transportation System (ITS), loss of traffic data seriously influences the accuracy of decision-making of urban traffic planning and management. To solve this vital challenge, we introduce a new algebraic framework for tensor decompositions with different representation of tensor rank for traffic missing data imputation. We propose a novel tensor completion algorithm by using tensor factorization and introduce a spatial-temporal regularized constraint into the algorithm to improve the imputation performance. The simulation results with real traffic dataset demonstrate that the proposed algorithm can significantly improve the performance in terms of recovery accuracy compared with other tensor completion algorithms under different data missing patterns at all data loss rates. This also indicates that the proposed algorithm is more efficient for missing traffic data imputation by exploiting such an algebraic framework than the traditional multilinear algebraic framework for tensor decompositions.